Private Ownership and the Cost of Debt: Evidence from the Bond Market

Brad A. Badertscher

University of Notre Dame

Dan Givoly* Penn State University

Sharon P. Katz

Columbia University

Hanna Lee University of Maryland

October 15, 2014

Preliminary--Please do not cite or distribute without the authors' permission

ABSTRACT

Using a sample of public bonds issued by privately-owned and publicly-owned companies we find that, after controlling for financial fundamentals and information environment effects, the cost of public debt issued by privately-owned companies as captured by ratings and yield spreads is significantly higher than that issued by publicly-owned companies. This higher cost, however, is justified, but only in part, by higher than expected actual rates of default among privately-owned firms. Among privately-owned companies, the cost of debt is higher for companies controlled by private equity (PE) firms. However, ownership by large PE firm reduces the cost of debt to their investees as compared to those owned by smaller PE firms. The results contribute to our understanding of the costs of public versus private ownership and our knowledge on the role of ownership type and “soft” information in bond valuation.

*Corresponding Author: Phone: 814-865-0587; E-mail: dgivoly@psu.edu Keywords: public firms, private firms, cost of debt, bond valuation, debt default risk, and bankruptcy JEL classification: G30, G32, G33, M41 Data Availability: Data are available from sources identified in the paper. ______________________________________________________________________________We are grateful for constructive comments from Maria Loumioti. We would like to thank PricewaterhouseCoopers, the Mendoza College of Business and the Robert H. Smith School of Business for financial support.

1

Private Ownership and the Cost of Debt: Evidence from the Bond Market

1. Introduction

A considerable body of analytical and empirical research examines the effect of

ownership structure on firm value. Among the value-driving ownership characteristics

investigated by past research are concentration, disparity between control and cash flow

rights, identity of the controlling group (management, employees, or family) and the type

of ownership - public or private.

With respect to the valuation effects of public ownership, various characteristics of

public ownership relative to private ownership that have bearing on the firm’s value have

been identified, the most obvious of which is the greater liquidity of public equity

ownership. The net effect of these characteristics on the cost of public equity appears to

be favorable, as evidenced by the so-called “public equity premium” or, conversely, the

“private equity discount.”

The effect of ownership type on the cost of debt, however, is less clear. On one hand,

the greater ability of the publicly-owned firms (hereafter, public firms)1 to raise equity

capital, the typically richer and more transparent financial information that they provide,

and their lower ownership concentration may result in a lower cost of debt. On the other

hand, the more restrictive regulatory environment of public firms, the myopic behavior

and opportunistic reporting of their management induced by equity-based compensation

and capital markets’ pressures, and their greater exposure to litigation risk (Badertscher

Jorgensen, Katz and Kinney 2014), may lead to a higher cost of debt. The extent to which

1 For ease of exposition, we refer to firms with private equity and public debt as “private” firms while firms with public equity and public debt are referred to as “public” firms.

2

these opposing factors affect the cost of debt of private as compared with public firms

remains an open empirical question.

In this paper we attempt to fill this gap. In doing so, we face two difficulties. First,

financial information on privately-owned companies, at least in the U.S., is limited to

regulated firms (primarily in the insurance and banking industries). Second, it is difficult

to disentangle the effect of ownership on the cost of debt from that of the reporting and

disclosure regime that is much stricter for public companies.

We alleviate these empirical difficulties by comparing the cost of debt of publicly-

owned companies to that of privately-owned companies considered “public” by the virtue

of issuing debt to the public. Even though privately-owned, these companies are subject

to the same financial reporting and disclosure requirements as their publicly-owned

peers.2

We estimate the difference in the cost of the public debt between publicly- and

privately-owned issuers. To isolate the effect of ownership factor on the cost of debt, we

control for all other determinants of cost of debt proposed in the literature. We also

control for differences that still exist between these two groups of firms in the quality and

transparency of information as well as management characteristics (such as

concentration).

Our main sample consists of 256 private equity and public debt (hereafter “private”)

firms and 3,415 public equity and public debt (hereafter “public”) firms, which represent,

respectively 1,150 and 29,193 firm-years over the years 1987-2010. For some of our tests

2 The definition of “security” in the Securities Act of 1933 and the Securities Exchange Act of 1934 includes both stocks and bonds.

3

we use an additional sample of 166 firms (represented by 670 private firm-years and 959

public firm-years) that change their ownership during the sample period from private to

public or vice versa. We find that, after controlling for firm and bond characteristics

identified by past research as affecting the cost of debt, the extent and quality of

information available on these firms, as well as the endogenous nature of the ownership

choice, the cost of debt, measured alternately by yield spread and rating, is higher for

private firms.

We further show that private firms experience a higher frequency of distress events

relative to what would be expected based on their fundamentals.3 This greater default risk

to creditors of private firms is likely related the more limited access of these firms to the

public equity market. Yet, these higher default rates should be reflected already in the

rating. Yet, when we control for rating, the spread for private bonds is still higher by

about 1.4% than the spread on comparable public bonds. This unexplained discount of

private bonds suggests that investors over-discount these bonds.

In addition to comparing the cost of debt between public and private firms, we also

compare this cost between different types of private firms. Specifically, within private

firms, we find that ownership by private equity (PE) firms is associated with a higher cost

of debt, consistent with the notion that greater separation of ownership and control in PE-

backed companies as compared to other private firms, may lead to more risk taking

(Fama and Jensen 1983, Badertscher, Katz and Rego 2013).4 Similarly, majority-

3 We follow Moody’s definition of default which includes distressed exchanges, Chapter 11 and Chapter 7 bankruptcies, dividend omission, and missed interest and/or principal payments.

4 We note that these results are in contrast to the findings of lower cost of syndicated loans obtained by PE-backed companies (e.g., Demiroglu and James 2010; Ivashina and Kovner 2011; and De Fontenay 2013).

4

ownership by PE firms leads to higher cost of debt as compared to minority-ownership by

PE-backed companies. We further show that ownership by large PE firms is associated

with a lower cost of debt to their investees as compared to ownership by smaller PE

firms, consistent with their greater reputational concerns as repeated players in the capital

markets.

The paper makes a number of contributions to the literature. The finding that the net

favorable effect of being public on the cost of equity capital extends to the cost of debt

contributes to our understanding of the consequences of the ownership choice on the

firm’s cost of capital. The findings concerning the influence of ownership type on bond

ratings extends the literature on the role of “soft information” in the determination of

credit ratings. The paper further extends the literature on the effect of ownership

characteristics and the identity of its major shareholders (e.g., PE firms) on the cost of

debt. Lastly, our finding contributes to our understanding of the effect of ownership

structure on default risk.

The remainder of the paper is organized as follows. The next section reviews the

literature on ownership characteristics and their expected effect on the cost of capital in

general and the cost of debt in particular. Section 3 presents and discusses the hypotheses.

Section 4 describes the empirical design. Section 5 describes the data and sample. Section

6 presents the results and their discussion. Robustness tests are described in section 7, and

concluding remarks are provided in the last section of the paper.

2. Literature Review

Early finance theory viewed stockholders as a dispersed yet homogenous group, and

management as an agent that acts in the best interests of the stockholders (McConnell and

5

Servaes 1990), being disciplined by either takeover threats (Manne 1965; Martin and

McConnell 1990) or the labor market (Fama 1980). This view of the irrelevance of

ownership structure and the benign behavior by management has evolved and altered

over time with the development of agency theory which introduced complexity and

richness into the management-ownership relation. In their seminal paper, Jensen and

Meckling (1976) show formally how agency costs arising from the separation of

ownership and control (management) affect firm value. Subsequent work has dealt with

the effect on firm value of various ownership and management characteristics such as the

separation of ownership and control, management ownership, ownership concentration

and the identity of the major equity holders (e.g., family, the public, institutional

investors, PE firms, etc.)

2.1. Theory and Evidence on the Effect of Ownership Features on the Cost of

Debt

2.1.1. Ownership concentration and disparity between ownership and control

The disparity between ownership and control has been shown theoretically to lower

the firm’s value (Grossman and Hart 1988; Harris and Raviv 1988). The empirical

evidence is consistent with this prediction. For example, Classence et al. (2002) show that

the beneficial incentive effect of having a large number of shareholders dissipates and, in

fact, become negative when the extent of control exerted by them exceeds their cash flow

rights. Aslan and Kumar (2012) present a model where raising the dominant

shareholder’s ratio of control-to-cash-flow ownership increases the unconditional

probability of default and lowers the creditors’ payoffs conditional on a default. The

equilibrium cost of debt is therefore positively related to the disparity between control

6

and cash flow rights. The evidence they provide is consistent with their model’s

prediction. Relatedly, Johnson et al. (2000), Gilson (2006) and Jiang, Lee, and Yue

(2010) provide evidence that ‘tunneling’ by the dominant shareholders who control the

firm adversely affect debt-holders.

2.1.2. Public versus private ownership

Public firm’s equity, primarily because of its liquidity and the easier access of the

public firm to the capital markets, is traded at a premium relative to the value of a similar

private equity.5 The more limited access of privately own firms to the equity market is

likely to affect also their cost of debt. To the extent that private firms’ equity shares have

lower liquidity compared to those of public firms, private firms may be more likely to

become financially constrained or distressed when their operating performance

deteriorates, all else equal. Therefore, public debt market may require a higher premium

for private firms to compensate for higher ex ante credit risk private firms can have.

Other characteristics along which these two types of equity ownership differ and

which are likely to affect the cost of debt include the extent of separation between

management and ownership, the degree of regulation, the level of litigation risk and the

structure of management compensation. These characteristics affect the firm’s cost of

debt either directly or indirectly through their effect on the disclosure and reporting

attributes of these two types of equity ownership (see, for example, Ball and Shivakumar

2005, Burgstahler et al. 2006, and Givoly et al. 2010 for evidence of this effect).

5 The extent of this public equity premium (or private equity discount), however, is subject to debate (see for example, Hertzel and Smith 1993, Koeplin et al. 2000, Das et al. 2002, Kooli, et al. 2000, and Comment 2012). See Bruner et al. (1998) for a review of the different approaches to measuring the premium.

7

To the extent that private firms’ ownership characteristics exacerbate or ameliorate

agency problems, such characteristics will be associated with a higher or lower cost of

debt. On the one hand, the weaker separation of ownership and control in privately

owned firms (Badertscher et al. 2013) reduces agency problems between the management

and stakeholders, resulting in less severe agency problems for these private firms. On the

other hand, private firms have more concentrated ownership and control, leading to more

acute conflicts between shareholders and debt-holders. Taken together, the net effect of

private ownership on the cost of debt due to agency conflicts is ambiguous.

Other private ownership characteristics, however, have a clearer directional effect on

the cost of debt. First, private firms have limited access to the equity capital markets,

which imposes restrictions on external financing. Because of the limited access to capital,

private firms are more likely to forego positive net present value investments (Stein 2003;

Lin et al. 2011) and more likely to default.6 While not completely offsetting these costs,

there are benefits of being private in the form of a lower risk of litigation (Badertscher et

al. 2014) and less disperse ownership that facilitates debt renegotiation or restructuring

that, in turn, reduce the probability of default.7

A second characteristic of private ownership that is likely to increase the cost of debt

to private firms is that these firms operate in a more opaque information environment, are

covered less by the press, have limited analysts’ coverage, and do not have stock prices

available (Katz 2009). These traits render private firms riskier in the eyes of public debt-

6 Ivashina and Scharfstein (2008) find evidence that firms are more likely to default when liquidity deteriorates. Campello, Graham, and Harvey (2010) provide survey evidence on the importance of liquidity when a firm is financially constrained or is in distress. 7 Gilson, John, and Lang (1990) provide evidence that firms with more complex capital structure are more likely to fail in private restructuring, hence have enhanced probability of filing for bankruptcy.

8

holders who assign these firms a higher default risk, especially as this higher level of

information asymmetry exacerbates moral hazard problems between management and the

stakeholders (Jensen and Meckling 1976). Several recent studies examine empirically the

cost-of-debt differential between private and public companies. Pagano et al. (1998)

examine the change in the cost of debt of 40 Italian IPOs and find a reduction in that cost

following the IPO. However, their design does not allow determining whether the cost

reduction is a result of a change in ownership or of improved public information

associated with stock exchange listing. Saunders and Steffen (2011) test the cost of

syndicated loans for privately held vs. publicly traded companies in the United Kingdom

and conclude that privately held firms incur a higher cost, after controlling for borrower

and loan characteristics. They attribute the difference to the higher cost of information

production associated with private firms. Note that Saunder and Steffen (2011)’s main

analysis is conducted on private firms with private debt. These firms are not subject to the

additional disclosure and auditing requirements of the stock exchange. Our sample

consists of public firms with a different type of ownership (all subject to the same

reporting and disclosure requirements). As such it enables us to better separate the

information from ownership explanations. Saunders and Steffen (2011) do examine

within their sample of private companies the effect of sub-types of private ownership on

the cost of debt. In particular, they compare the cost of loans between private firms with

public debt and public firms, but do not find significant differences. We note, however,

that syndicated loans have different attributes than public bonds. For example, the banks

that syndicate the loans benefit from private information and seniority of their claims.

9

Similar to our study, Kovner and Wei (2012) focuses on the cost of public debt of

private vs. public firms. They find that bonds issued by private firms are discounted

relative to those issued by public firms. Their study differs from ours in a number of

respects. First, it tests the bond pricing at issuance. This is a time in which the private-

firm-turned-public has an obvious informational disadvantage relative to its more

established public companies, making it difficult to attribute the cost-of-debt differential

to the ownership type.8 In addition, in assessing the effect of ownership on the cost of

debt, Kovner and Wei (2012) control for the effect of firm’s fundamentals using a limited

set of variables that includes only total assets, total debt, and EBITDA, making the

attribution of the results to ownership type less reliable.9 Further, they do not distinguish

between different ownership characteristics of private firms (e.g., ownership by PE

firms).

2.1.3. Private ownership through a private equity firm

Equity may be held privately by an investment company denoted as a “private

equity” (PE) firm. PE-backed private companies have certain ownership attributes that

have a bearing on their cost of capital. Typically, PE firm raises pools of capital from

institutional investors and acquires majority control of mature, profitable businesses via

leverage buyouts (LBO) with the objective of holding them for five to seven years while

improving their financial performance and value. The return on these investments is

8 In robustness test, they examine the secondary market’s pricing. However, that sample is limited to 40 bond issues. 9 The limited set of control variables is apparently due to missing 10K information for many of the sample firms in Kovner and Wei (2010). Missing 10K information would occur for public companies with less than 500 stockholders for which 10K filing is not required. The presence in the sample of these firms, for which limited public information is available, makes it likely that the cost-of-debt differential attributed to ownership type stems in fact from another source – the extent and quality of information.

10

typically in the form of capital gains from an IPO, a merger, or the sale to another

company.

The partners of the PE firms, while not involved in the active management of their

investees, sit on the boards of the private companies in their portfolio and are quite active

in advising these companies and monitoring and motivating their management teams

(e.g., Kaplan and Stromberg 2009; Masulis and Thomas 2009).

The effect of the above characteristics of ownership by PE firms on their investees’

cost of debt is not clear. PE-backed private companies typically have greater separation

of ownership and control which may lead to more risk taking (Fama and Jensen 1983,

Badertscher et al. 2013) and thus likely result in a higher default risk, which translates to

higher cost of debt. On the other hand, PE firms, being repeat players in the capital

market, are likely to use their reputation with creditors to mitigate the problems of

borrower adverse selection and moral hazard associated with lending to private

companies. This could enable PE-backed companies to obtain lower cost loans and

benefit from less restrictive debt covenants (see Demiroglu and James 2010; Ivashina and

Kovner 2011; and De Fontenay 2013). These positive reputational effects of PE

ownership are a function of the strength and market position of the PE firm and therefore

more likely to be pronounced when the PE firm is large in terms of assets under

management. The proportion of stock owned by top executives at minority-owned, PE-

backed companies, is significantly greater than managerial stock ownership at majority-

owned, PE-backed companies (Katz 2009, Badertscher et al. 2013). Therefore, the

separation of the ownership and control, and its effect on risk taking, is more likely to be

pronounced when the PE firm holds majority-ownership in its portfolio firm.

11

3. Hypotheses

The findings by prior research on how firm ownership might influence the cost of

debt are mixed. Nonetheless, the preponderance of the situations in which that cost of

debt would be higher for private companies (see the discussion in section 2.1.3 above)

leads us to the following directional hypothesis:

H1: The cost of debt is significantly higher for private firms than for public firms.

As the review of the salient literature indicates, each type of ownership structure

within the private firms (i.e., ownership by a PE firm) has characteristics that have

conflicting effects on the cost of debt. Therefore we have no prediction about the

direction of their effect on the cost of debt and, accordingly, test the following non-

directional hypotheses:

H2: The cost of debt of private companies is not significantly different from that of

other private companies in which a PE firm owns a share.

H3: The cost of debt of private companies in which a PE firm is a shareholder is not

significantly affected by the size of the PE firm.

H4: The cost of debt of private companies in which a PE firm is a shareholder is not

significantly affected by PE firm majority- versus minority-ownership.

4. Research Design

4.1. Measures of the Cost of Debt

We use two alternative measures that capture the cost of debt through the underlying

credit risk of the debt security: the credit rating (RATING) and the yield spread

(SPREAD). These measures are assessed once a year, at the beginning of the fourth

month of the fiscal year.

Credit rating, RATING, is based on S&P issuer-rating system measuring quality of

the credit on a firm-level. For the statistical analysis, we convert the rating designations

12

from AAA (the highest) to D (the lowest) into numerical scale, ranging from 1 (AAA) to

21 (D).10 The definitions of all of the variables used in this study are provided in

Appendix A.

SPREAD is calculated by subtracting from the yield-to-maturity of the company’s

bond-year the concurrent yield-to-maturity of matched Treasury bonds. We match each

bond-year with a Treasury bond in the CRSP database that has (1) the same remaining

time to maturity in years as the corporate bond (specifically, a Treasury bond that

matures no more than six months before or after the time to maturity remaining of the

corporate bond) and (2) the closest annual coupon rate. We further require that the ratio

of the absolute difference between the coupon rate on the corporate bond and the coupon

rate on the matched Treasury bond not exceed 0.45 of the corporate bond’s coupon rate.

This requirement results in about 10% of the bond firm-year observations not having

SPREAD information.

4.2. Determinants of Cost-of-Debt

In assessing the effect of firm ownership on the cost of debt, we control for other

determinants of the cost-of-debt. The determinants used in our analyses are based on

prior research (for a survey of this research, see Ashbaugh-Skaif et al. 2006). They

consist of two sets of variables. One set is that of firm fundamentals and bond

characteristics, most of which are credit risk-related. The other set of variables is

designed to capture the information environment of the firm in terms of the extent and

10 While this numerical conversion has been used by other studies (e.g., Amato and Furfine 2004, Kisgen 2006), we also used, as part of our robustness tests, alternative non-linear conversions with no effect on our inferences.

13

quality of information about the firm. All of the determinants are measured for each firm-

year. The set of firm fundamentals is represented by the following variables:

(1) LEV_ADJ: Leverage (the ratio of total liabilities-to-total-assets) adjusted for

off-balance sheet accounts,11

(2) INT_COV: The interest coverage ratio computed as operating income before

tax divided by interest expense (the value is set to zero when the numerator is

negative,

(3) Z SCORE: The Altman Z-Score (Altman 1968) is computed using updated

weights of its core factors as derived by Shumway 2001. The measure consists

of five components - working capital, retained earnings, and EBIT, all deflated

by lagged assets, the sales-to-assets ratio and the ratio of the market-value-of-

equity- to- debt. Following Shumway (2001), in our tests we use only the first

four factors since there is no market value of equity for private firms,

(4) ROA: Return-on-assets computed as the ratio of income from continuing

operations to lagged total assets,

(5) LOSS: An indicator variable that is set equal to one if the firm reports

negative earnings in the current fiscal year,

(6) FCF: Free cash flow computed as cash flow from operations less average

capital expenditures in the most recent three years, deflated by lagged total

assets,

(7) INTANG: The ratio of intangible assets to total assets,

11 LEV_ADJ is computed as [(Total Liabilities + Underfunded Pension) + (rent expense *6*(1-τ))]/[Total Assets + τ * Underfunded Pension + LIFO Reserve * τ + Interest Capitalized * (1- τ)) + (Rent Expense*6*.(1- τ))], where τ is the statutory corporate tax rate (0.35 in our sample period) – see Moody’s 2007.

14

(8) PPE: Capital intensity computed as the gross balance of property, plant and

equipment deflated by total assets,

(9) GROWTH: Sales growth calculated as the change in sales deflated by lagged

sales,

(10) DURATION: Average duration of the firm’s outstanding bonds expressed as

a number of years,

(11) SENIORITY: An indicator variable that is set equal to one if the firm has a

senior bond, and zero otherwise.

(12) FSIZE: Firm size, measured by natural log of the sum of the market value

of the equity (imputed market value for private firms)12 and the book value of

debt. Other fundamentals being equal, large firms are more likely than smaller

firms to access more resources to avoid default than small firms. Large firms

are, further, more likely to be diversified than small firms, which reduces the

uncertainty of their future cash flows.13 At the same time, firm size, firm size

relates to the information environment in which the firm operates: Because it

represents the market value of the firm, it is related to analyst following and

institutional ownership, thus indicating greater research efforts and breadth of

information about the firm. Larger firms also are subject to a greater political

cost, which tends to improve the quality of their reporting. For these reasons,

12 The market value of the equity of privately owned firms is estimated by multiplying the firm’s assets, and separately, the firm’s sales, by the median in the firm’s 3-digit industry in the year of, respectively, the assets and sales multipliers, and averaging the two products. 13 We also use an alternative measure of size, the dollar value of the bond issue at the time of its initial offering. This variable, which is highly correlated with FSIZE yields similar results.

15

firm size may be viewed as a hybrid variable, conveying both fundamentals

and information environment factors.14

To further explore the effect of the information environment in which the firm

operates, we also examine two additional explanatory variables beyond firm

fundamentals and firm size:

(13) BIGAUDI: An indicator variable that is equal to one if the firm’s auditor

belongs to one of the Big-5 auditors (Big-4 after the demise of Arthur

Anderson in 2002) and zero otherwise. This variable is designed to capture the

quality of the firm’s financial reporting (see Mansi, Maxwell, and Miller

2004).

(14) PUBLIC NEW: The length of time the firm has been public. It is captured by

an indicator variable that is equal to one if the firm has been public (either debt

or equity) less than five years and zero otherwise. A longer history of the firm

as a public corporation is likely to be associated with availability of more

public information and greater familiarity of investors and creditors with the

firm’s operations, management, and prospects.

The first hypothesis of the paper concerning the cost of debt of public and private

firms, H1, is tested through the following cross-sectional regressions estimated over

pooled firm-years, using alternately RATING and SPREAD as a measure of the cost of

debt (firm and year subscripts are omitted):

SPREAD/ RATING = f {PRIVATEt, Determinants of Cost-of-Debt} (1)

14 We also use an alternative measure of size, the dollar value of the bond issue at the time of its initial offering. This variable, which is highly correlated with FSIZE yields similar results.

16

SPREAD and RATING are defined and discussed in the previous section and the

determinants of the cost of capital are presented above. Our variable of interest in

regression 1 is PRIVATE, an indicator variable, receiving the values of one if the firm is

privately owned and zero otherwise. A significant positive coefficient on this variable

would indicate that, after controlling for risk-related firm fundamentals, the cost of debt

of private firms is higher than that of public firms. All regressions further include

industry and year fixed effects. With appropriate changes, regression (1) is also used to

test hypotheses H2, H3 and H4 that pertain to the effect of the identity of the major

shareholders of the firm (e.g., PE firms) on the cost of debt.

4.3. Likelihood of Default and Rate of Recovery

To facilitate the interpretation of the main results regarding the effect of private

ownership on the cost of debt, we also examine whether private ownership affects the

two parameters that determine credit risk, namely, the likelihood of default and the extent

of recovery by the creditors of their debt to the firm in case of a default. For the default’s

examination, we employ the hazard model suggested by Shumway (2001).15 We augment

their model by additional measures of firm fundamentals suggested by the literature that

are used as determinants of cost of debt in regression (1). The hazard model is tested

using the following cross-sectional regression estimated over pooled firm-years (firm

subscripts are omitted):

Pr (DEFAULTt) = f {PRIVATEt-1, LEV_ADJt-1, INT_COVt-1. ZSCOREt-1, ROAt-1, Losst-1, FCFt-1,

INTANGt-1, PPEt-1 , GROWTHt-1, FSIZEt-1} (2)

15 Shumway 2001 and Campbell, Hilscher and Szilagyi 2008 provide evidence that hazard rate model based on reduced form econometric specification outperforms other distress prediction models (e.g. discriminant analysis, simple logit) in terms of both in- and out-of-sample forecasting accuracy.

17

where a DEFAULT event (0,1) is defined as either bankruptcy (Chapter 11 and Chapter 7

bankruptcies), default (missed interest and/or principal payments), distressed exchange16

or dividend omission.17 All of the independent variables are as defined in section 4.2.

Following Shumway (2001), we estimate the hazard rate model in the equation

(2) by employing multivariate logit approach to estimate the effect of private ownership

on the default risk after controlling for accounting fundamentals.18 A significant positive

coefficient on PRIVATE would indicate that, after controlling for default risk-related firm

fundamentals, the ex post default likelihood of private firms is higher than that of public

firms, which would be consistent with private firms having higher default risk than public

firms.

The other parameter that determines credit risk is the rate of recovery. To test

whether private ownership affects the recovery rate by the creditors in the wake of a

bankruptcy or other default, we estimate a regression model following Acharya, Bharath

and Srinivasan (2007) and Donovan, Frankel, and Martin (2013). Specifically, the

expected rate of recovery is estimated using the following regression model:

Recovery Rate = f {PRIVATE t-1, EBITDA t-1, TA t-1, TAN t-1, LEV_ADJ t-1, DUR t-1, DE t-1,

SECURED t-1} (3)

16 “Distressed exchange” refers to a fundamental change in the contractual relationship between a debtor and its creditors such as a reduction in the effective interest rate, extension of time to repay, subordination of claims, or substitution of lower priority equity securities for debt claims. 17 We follow Moody’s definition of dividend omission that is characterized as a default event by Moody’s. 18 Shumway 2001 analytically shows that hazard rate model can be estimated by multivariate logit approach including all the available panel data of firm-year observations. Our results employing the logit approach, however, are unaffected when we use proportional hazard model (Cox 1972).

18

where Recovery Rate is defined as Moody’s family recovery rate from Moody’s DRS

database, and PRIVATE (an indicator variable), EBITDA,TA, and LEV_ADJ are defined

in section 4.2 above. TAN is tangibility of the assets defined as the fraction of total assets

represented by Property, Plant and Equipment. DUR is the duration of distress event

measured as the difference in the number of months between the date of distress event

and the date of resolution of distress from Moody’s DRS database following Donovan et

al. (2013). Donovan et al. (2013) include DUR in their recovery rate regression to

account for the intuition that the longer the duration of distress, the more likely that lower

the rate of recovery from distress. DE is an indicator variable that is 1 if the distress event

is a distressed exchange and 0 otherwise. This variable is designed to capture higher

recovery rate of firms going through distressed exchanges compared to other types of

default events such as bankruptcy shown by Franks and Torous (1994). Lastly,

SECURED is an indicator variable that is 1 if the firm has secured debt, and 0 otherwise.

If the firm has secured debt at the time of default, secured creditors are likely to have

higher recovery rate everything else the same. Therefore, we include this variable to

capture cross-sectional variation in recovery rates driven by secured status of debt

claims.19

As with the case of the distress prediction model regression (2), the variable

PRIVATE is introduced in order to assess, in this case, whether, after controlling for

19 Donovan et al. 2013 use proportion of secured debt as a determinant of the recovery rate. While this measure is likely to be a sharper measure of the percentage of distressed claims that are secured, measurement of this variable relies on Moody’s DRS database, which substantially limits the sample. We employ an alternative definition by collecting sample firms’ secured debt status from Moody’s DRD as well as Mergent FISD and SDC database and hand-matching them to our database of private-public firms with the recovery rate information. However, our results are qualitatively the same when we apply the alternative definition of SECURED.

19

firm’s characteristics that affect the rate of recovery, that rate is also influenced by

whether the company is private or public.20 Following Acharya et al. 2007, determinants

of recovery rates are measured at t-1, the firm-year immediately preceding the distress

event.

4.4. Construction of a Matched Sample

Regression (1) is estimated from a pooled sample of firm-years across firms and

years. As part of our robustness tests we also estimate the regression from a matched

sample of firm-years. The matched sample is constructed by matching each private firm-

year with a public firm-year. The matching is based on the proximity of the matched firm

to the test firm’s score (i.e., propensity score). We match each of the firm-year bond

observations belonging to a private-equity firm to a firm-year bond observation

belonging to a public-equity firm. In order to reduce the problem of endogeneity, we

identify the determinants of the choice of being a private company, based on the findings

of past research21 and estimate the coefficients from a regression of the ownership type

(the PRIVATE variable). The dependent variable in the propensity score’s logit model is a

private-equity firm indicator variable (PRIVATE) and the independent variables reflect

firm characteristics such as size, financial risk and constraints, growth opportunities and

asset composition. Specifically, the model is estimated and scores of potential matches

among public firms are determined by using the following independent variables:

INT_COV, LOSS, INTANG, PPE, and GROWTH (all defined in section 4.2. above).

Because of the presumed strong influence on the choice of ownership type of industry

20 Following prior studies, we also estimate regressions (2) and (3) using LEV instead of LEV_ADJ, with essentially the same results. 21 See, for example, Ball and Shivakumar 2005 and Givoly et al. 2010.

20

affiliation, firm size and leverage, for the public firm-year that is matched we imposed a

further restriction on the selection of the matched sample such that, in addition to having

the closest score to a given private firm-year, the selected firm-year observation must

belong to the same fiscal year, 3-digit industry, quintile of total assets’ distribution and

quintile of leverage distribution as the private firm-year. Finally, to ensure that each

private-equity firm-year and its match are reasonably similar to each other, we restrict the

two firms to have propensity scores within 0.10. We allow a public-equity firm-year to

serve as a match only once per year.

5. Sample and Data

5.1. Sample

Our sample consists of firms with public debt in the 24-year period 1987-2010. To

identify private firms that have publicly-traded debt we follow the procedure used by

Katz 2009, Givoly, Hayn and Katz 2010 and Badertcher et al. 2013. Specifically, we

select on COMPUSTAT in any of the sample years firm-years that satisfy the following

criteria: (1) the firm’s stock price at fiscal year-end is unavailable, (2) the firm has total

debt as well as total annual revenues exceeding $1 million, (3) the firm is a domestic

company, (4) the firm is not a subsidiary of another public firm and (5) the firm is not a

financial institution or in a regulated industry (SIC codes 6000-6999 and 4800-4900).

We exclude financial institutions (SIC codes 6000-6999) from our sample since many of

the accounting items and financial ratios used in our analyses do not apply to these firms.

We also exclude from our main sample firms that changed their ownership during the

sample period, denoted as “transitioning” firms. This sub-sample is tested separately, as

21

explained below. Finally, we exclude from consideration convertible and callable bonds

because of their more complex valuation.

Of the sample of firm-year observations belonging to the firms that meet the above

criteria, we retain only those firm-years for which information on at least one of the cost-

of-debt measures (SPREAD or RATING) is available.

The resulting main sample consists of 1,150 firm-years of private firms representing

256 distinct firms and 29,193 firm-years of public firms representing 3,415 distinct firms.

Our tests are conducted on both the full sample and the propensity-matched sample. The

latter is constructed through the process described in section 4.4 consists of 343 firm-

years and their distinct matched public firms.

We also take advantage of the sample of “transitioning” firms by conducting a

supplementary test of the main hypotheses based on the firms’ time-series. This sub-

sample consists of 73 firms that were initially private (representing 268 yearly

observations as private firms) and have subsequently become public (resulting in 397

yearly observations as public firms) and 93 that were initially public (representing 402

yearly observations as public firms) and have subsequently become private (resulting in

562 yearly observations as private firms).

Some firm-years contain more than one bond outstanding. About 17% of the firm-

years of private firms and about 42% of the firm-years of publicly owned firms contain

more than one outstanding bond (most commonly two). In all, our sample of firm-years

reflects 36,771 separate bond-years, or an average of 3.06 bonds per firm-year. To avoid

giving undue weight to firm-years with multiple bond issues outstanding and to avoid

cross-sectional dependence of the observations, each firm with more than one outstanding

22

bond issue in any year is represented in the statistical analyses only once for that year.

This is accomplished by averaging the SPREAD and RATING for that firm-year across

the firm’s bond issues outstanding in the year.

For the purpose of testing hypotheses 2-4, we further constructed subsamples of

private firms based on the identity of their owners. Specifically, we categorized private

firms as being PE-owned, defined as firms whose equity is held in part by PE firms (with

further partitioning of these subsamples into majority- and minority PE ownership and

small and large PE owners).

5.2. Data

Monthly bond price data is obtained from Interactive Data Pricing and Reference

Data, a provider of third-party bond prices and other financial services, whose subscribers

include thousands of financial institutions worldwide ranging from central banks to large

investment banks.22 In collecting bond price data, Interactive Data prioritizes its data

sources, reporting transaction-based bid prices when available and using either

institutionally-based matrix bid prices or dealer bid quotes (referred to as “evaluated

prices”) to fill in the series for periods where bond bid prices are missing (generally as a

result of infrequent trading). Bond prices, spreads and duration for the years 2008-2010

are obtained from TRACE.

Financial statement information and S&P rating information is obtained from

COMPUSTAT. Data needed to determine ownership and ownership rates of private

equity firms by PE, or management are hand-collected from SEC filings. We construct

22 Other research using this database includes Hemler (1990), Gay and Manaster (1991), Hand et al. (1992), Shulman, and Bayless (1993), Cooper and Shulman (1994), Hancock and Kwast (2001), Dudney and Geppert (2008), and Givoly et al. (2013).

23

the following ownership variables based on these data: Ownership by a PE firm (PE),

majority ownership by a PE firm (PE_MAJ), and ownership by one of the largest PE

firms in terms of dollar investments (PE_RANK). All of these variables are defined in

Appendix A.

Information on distress events is based on the database constructed by Lee (2014).

That database consists of bankruptcies (Chapter 7 and Chapter 11 bankruptcies), defaults,

distressed exchanges, and other default events following Moody’s definition of default.

That database contains 4,897 non-overlapping distress events of non-financial firms for

the period 1980-2011.23 Matching the default events to the private-public database used

in our study yields a sample of 879 firm-years containing default events. Missing

financial data needed for distress prediction reduces this sample to 738 firm-years with

111 private firm-years and 627 public firm-years. Together with the public-private

database, we have 32,726 non-distress firm years, and 738 distress firm-years to conduct

our hazard rate tests.

We obtain firm-level data on the rates of recoveries from Moody’s Default Risk

Service (Moody’s DRS) database, one of the most comprehensive sources of debt- and

firm-level credit recovery rates data for firms. We match the firm-level recovery rate

(denoted by Moody’s as “family recover rates”) with our private-public-firm database

combined with cost of debt data. Our final sample of recovery rates consists of 374 firm-

years with 56 private firms and 318 public firms.

23 The database is compiled and hand-matched from a number of sources including FACTIVA, Lexis Nexis, Capital IQ, PACER, SDC Platinum restructuring database, Moody’s rating database, the CRSP Monthly Stock file, the website Bankruptcy.com, and the list of bankruptcy filings generously provided by Lynn Lopucki.

24

5.3. Descriptive Statistics

Table 1 shows descriptive statistics about the sample of firms and firm-years that is

used to test our main hypotheses. The panel shows that the distribution of the sample

firm-years over the years is not meaningfully tilted toward periods of economic

expansion (e.g., 2003-2007) or contraction (e.g., 2000-2001 and 2008-2009).

Panel B of table 1 shows the distribution of the sample firms by industry. Comparing

columns (5) and (6) reveals no strong over- or under- representation by the combined

sample of the population distribution of firms by industry. The comparison between

columns (2) and (3) shows in some industries there is greater representation by either

private firms (e.g., industry 27- Printing and Publishing) or public firms (e.g., industry

48) – Communications). To the extent that such disproportional industry representation

biases the results, this cause of bias is eliminated through the use of industry and year

fixed-effects and the construction of the matched sample.

Panel C of table 1 presents firm and bond characteristics for, the firm-years of,

separately, the private and public companies. It is apparent from the panel that the two

subsamples are quite distinct. Private firms are smaller, with FSIZE, defined as the sum

of the book value of the debt and the market value of the equity (imputed value, in the

case of private companies) being smaller among private than public companies, with a

median of $1.032 billion. (e6.939) vs, $2.527 billion of public companies. The private

firms are also less profitable (in terms of their return on asset (ROA) and frequency of

losses (LOSS). They further have a higher leverage and their asset composition is more

tilted toward intangibles (INTANG). Their Z_SCORE is lower (a median of 0.327

compared with a median of 0.612 of public companies), indicating a higher risk of

25

bankruptcy. The private firms also have a shorter history of having their securities

publicly traded (PUBLIC_NEW) with 53% of the firm-years are no more than five years

from the year in which the firm became public as opposed to only 16% of the public

firms.

The bonds issued by the private companies reflect their firm characteristics. Their

bond issuances are much smaller, a mean of OFFER of 5.025 (or $152 million) vs. 5.911

($369 million). The bonds issued by the private companies are riskier as captured by a

higher spread (a median spread of 7.9% vs. 1.9% of the public firms) and a lower rating

(a median rating of 14 vs. 10 for the public firms). All of the above differences between

firm and bond characteristics of private and public firms are statistically significant at a

higher than 1% level of significance.

6. Results

6.1 The Effect of Private vs. Public Ownership on the Cost of Debt

The main results on the effect of private ownership on the cost of debt are provided in

tables 2, 3 and 4. Table 2 shows the results from estimating regression (1) from the full

sample using, alternately the yield spread and rating as a measure of cost of debt. As

expected, all of the variables representing financial fundamentals are associated with both

the spread and the rating in the anticipated direction, and almost all of them are highly

significant.

Among the “information environment” variables, the variable FSIZE which is, as

explained in section 4.2., a hybrid variable representing both an economic fundamental of

the firm as well as the richness of its information environment, has a strong and very

26

significant effect on reducing the cost of debt. PUBLIC_NEW, which captures the age of

the firm as a public firm, indicates bonds of firms that have only recently become public.

The size of the audit firm as captured by BIGAUDI is insignificant, however. The impact

of the information variables on the cost of debt suggests that in setting the ratings and in

setting the spread, rating agencies and investors, respectively, take into account not only

the economic fundamentals of the firms as reflected in their financial statements but also

“soft” elements such as the information risk of the company, its management quality and

its accessibility to the capital markets.

The main variable of interest, PRIVATE is positive and highly significant, suggesting

that, after controlling for an array of fundamentals likely to affect the cost of debt as

measured by either SPREAD or RATING, private ownership is associated with a higher

cost of debt than public ownership. The fact that this variable is significant after

controlling for the information variables suggests that when investors price bonds, and

rating agencies decide on the ratings and investors set the yield, they consider other “soft

elements” beyond those represented by our “information environment” variables. An

alternative interpretation of the persistent significance of PRIVATE is that rating agencies

and investors overestimate the credit risk of private companies. We examine the validity

of the latter interpretation in the next section in which we compare the difference in

unexpected distress rates and rates creditors’ recoveries between debt issued by private

and public companies.

To gain an additional insight into the relative contribution of economic fundamentals,

information environment, and type of ownership to the explanatory power of the

regression, we decompose the R2 from regression (1), using Shapley’s decomposition

27

procedure (see Shapley 1953). The Shapley values indicate that of the total explanatory

power of the full regression that includes all of these three groups of variables (that

collectively yield an Adjusted R2 of 54.23% for the Spread regression, see table 2), firm

and bond fundamentals explains 62.4%, information variables 2.6% and type of

ownership 11.8% (industry and year fixed effects contribute 23.2% of the explanatory

power of the regression). This statistic highlights the fact that the ownership type has an

important effect on the cost of debt.

As explained in section 4.4, we conduct the analysis also on a propensity-score-

matched sample generated through a procedure that results in a sample of matched pairs

selected based on a propensity score procedure. The regression used to estimate the

propensity scores and the proximity of the matched sample to the test sample. The

estimates of the propensity-score regression are provided in the panel A of table 3; the

homogeneity of the private and public firm-years in terms of the variables underlying the

propensity-score estimation is shown in the second panel of the table. The table shows

that the propensity-score regression identifies correctly the variables associated with the

choice of the ownership type. All the variables, with the exception of LOSS, are

significant and the MacKelvey Pseudo R2 is 41.8%. Further, as panel B of table 3 shows,

the propensity-score process succeeds in generation fairly similar groups of public and

private companies: The differences in the variables underlying the propensity-score

estimation between public and private firms in the matched sample, which are significant

in the full sample (see panel C of table 1), are either no longer significant in the matched

sample or reduced in magnitude.

28

The results from estimating regression (1) obtained from the matched sample are

presented in table 4. The results are very similar to, and confirm those from the full

sample. In particular, the coefficient on PRIVATE continues to be positive and significant

for the matched sample. Further, the relative roles of fundamentals, information

environment and ownership type in explaining the cost of debt, as measured by the

Shapley values of these groups of variables (51.86%, 1.08% and 9.21%, respectively).

An important reason for the discount of bonds issued by privately owned companies

relative to those issued by public companies is the more limited access to the capital

markets that private companies have. The accessibility factor is expected to be more

influential at recessionary times for more financially constrained firms (Erel et al. 2012)

so we expect the ‘discount’ of bonds of private firms to be deeper in such periods. We

test this expectation by augmenting regression (1) by adding an indicator variable for a

recession year, REC, and an interactive variable REC*PRIVATE. The variable REC,

which receives the value 1 when the year is a recession year, as defined by National

Bureau of Economic Research (NBER), captures the fixed effect, if any, of recession year

on the mean spread or mean rating in our sample.24 The interactive variable captures the

additional ‘penalty’ for bonds issued by private companies in recession years as

compared with such discount in non-recession years.

The results, presented in table 5, shows no significant effect of recession years on the

cost of debt of the sample firms (REC is insignificant in both the SPREAD and the

RATING regressions). Yet, the excess of cost of debt of private vs. public firms in

24 Recession years in our sample period include 1990, 1991, 2001, 2008, and 2009 as defined by (NBER website: http://www.nber.org/cycles/cyclesmain.html.

29

recession years is significantly higher than in non-recess years. This result lends support

to the notion that limited access to capital markets in financially difficult periods is an

important explanation of the higher cost of debt of private firms. Further, the coefficient

of PRIVATE in the augmented regression is still positive and significant suggesting that

the higher cost of debt of private firms is not limited to recession years and thus appears

to not being fully explained by limited access to new capital only during recession

periods.

6.2. Effect of Ownership Structure on the Cost of Debt of Private Companies

As explained in sections 2 and 3, ownership of private companies by PE firms have

characteristics that can influence the cost of debt of these companies in either direction.

We first test the effect on cost of debt of PE ownership on the cost of debt. For this test,

we estimate regression (1) from the entire sample of private companies and replace the

variable PRIVATE with an indicator variable, PE, that receives the value of 1 if a PE firm

has some share in the ownership of the private company. We further test the effect on the

cost of debt of two such characteristics of PE ownership: The extent of ownership

(majority vs. minority) and the size of the PE. An indicator variable, PE_MAJ, receives a

value of 1 if the PE is a majority shareholder and 0 otherwise. We capture the size of the

PE by the total value of the investment portfolio held by the PE. The variable used to

measure the total value of the portfolio, is an indicator denoted as PE_RANK that

receives the value of 1 if the PE firm is among the top 15 PE firms by equity invested

(following Badertscher et al. 2013 and based on data obtained from Thomson Financial

VentureXpert) and 0 otherwise.

30

The results are shown in table 6. The variable of interest is DUMMY which is an

indicator variable that stands alternately for PE, PE_RANK and PE_MAJ described

above. The first two-column panel presents the results from estimating the augmented

regression from the entire sample of private companies. The DUMMY variable in this

panel, PE, is positive and significant, suggesting that having a private equity firm as an

owner increases the cost of debt of the private company, consistent with the greater

separation of ownership and control in PE-backed private companies, which may lead to

more risk taking (Fama and Jensen 1983, Badertscher et al. 2013) and thus likely result in

a higher cost of debt. The remaining panels of the table show results pertaining to the

effect on the cost of debt of the characteristics of PE ownership described above. These

results are obtained from variations of regression (1) estimated from the subsample of

observations belonging to private companies (about 2/3 of the observations on private

companies).

The first characteristic is the extent of ownership (majority vs. minority) by the PE

firm. Majority ownership accentuates ownership-control separation and thus leads a

higher cost of debt. Indeed, both coefficients on both spread and rating are positive and

the coefficient on spread is significant consistent with such higher cost of debt for this

subset of companies.

The second characteristic is the size of the PE firm in terms of value of its total

investment portfolio (PE_RANK). As explained in section 3, we expect that the positive

reputational effects of PE ownership more likely to be pronounced when the PE firm is

large in terms of assets under management. The coefficients of both variables are

negative and significant (for both the SPREAD and RATING regressions). This suggests

31

that private firms that are owned by large PE sponsors are more likely to use their

reputation with creditors to mitigate the problems of borrower adverse selection and

moral hazard associated with lending to private companies.

6.3. Ex-post Default and Recovery Rates

The finding so far consistently suggest that the cost of debt measured either by

yield spread or bond rating is higher for private firms, after controlling for both economic

fundamentals and the different information environment of these two groups of firms.

The most plausible explanation (but apparently not the only one -see the

discussion of the results of table 5 in section 6.1), is the more restricted access of private

firms to external financing. Such limited access should manifest itself ex-post in the form

of a higher default rate or a low debt-recovery rate than those expected given the firm and

debt fundamentals. We examine the validity of this explanation by determining whether

any of these rates is indeed different from its predicted values given the economic

determinants of each. For the purpose of this examination we predict distress events

(defined as bankruptcy, default, distressed exchange, or dividend omission), using a

variation of the hazard model proposed by Shumway (2001). The model, specified in

regression (2) also incorporates accounting-based predictors used by Altman (1968),

Zmijewski (1984), and Kovner and Wei (2012). To predict recovery rates we use to

model proposed by Acharya et al. (2007) and Donovan et al. (2013), as specified in

regression (3). The data on distress and recovery are described in section 5.2.

The results on the effect of private ownership on the probability of distress are

provided in table 7. These results are based on a sample of 30,877 observations (firm-

years) containing 738 distress cases. The table shows that all of the financial

32

fundamentals are significantly associated with realized distress outcome consistent with

the literature. The coefficient of PRIVATE is positive and significant, suggesting that the

distress rate is significantly higher for private companies, after controlling for firm and

bond fundamentals.

The assessment of the difference between private and public companies in

recovery rates following distressed events is based on 333 of the 738 distress cases for

which recovery data (from Moody’s DRS database) and accounting data needed to

estimate the model are available. The results, presented in table 8 shows that all the

fundamental predictors have the expected sign and most of them are significant. Our

variable of interest, PRIVATE has an insignificant coefficient, suggesting that the

recovery rate of creditor’s loans following a distress event is not different between

private and public companies.

6.3. Incorporation of Rating in the Explanation of Spreads

The above findings leave us with one explanation for the lower rating assigned by

credit rating agencies to private companies, namely, a higher rate of distress among these

companies, not captured by fundamentals, but likely related to the limited ability of these

companies to raise capital at times of distress. This conclusion is supported also by the

examination of the behavior of the cost of debt of private companies during periods of

economic recession (see section 6.1 above). It appears therefore that accessibility to

external capital is one of the “soft” factors considered by the credit rating agencies in

their rating decisions.

If credit ratings properly reflects the accessibility-to-capital factor, we would

expect that investors, who observed the rating, to incorporate this information as well.

33

That is, after controlling for RATING, the variable PRIVATE will no longer be associated

with the yield spread. To test this prediction, we re-estimated regression (1) with

SPREAD as the dependent variable, adding RATING to the set of explanatory variables.

Table 9 presents the results from this regression estimated separately from the full sample

and the propensity-matched sample. Both samples yield the same, somewhat surprising

result: While the magnitude of the ‘discount” of private debt is somewhat diminished

when RATNIG is added to the SPREAD regression, the variable PRIVATE is still

significant in explaining the yield spread after controlling not just for firm and bond

fundamentals as well as for information environment variables but also after considering

the bond rating.

A number of interpretations can be offered for this finding. One interpretation is

that credit rating agencies do not fully correct for the added risk of the debt issued by

private companies, yet investors (through the determination of the spread) correct for it.

Another interpretation is that the rating agencies properly rate credit risk of the private

companies but the market over-discount (“bashing”) private companies’ debt. A third

explanation for the finding is that there are important risk factors that are omitted from

our regression. Since we employ a broad range of explanatory variables that have been

tested by numerous past studies, we doubt that this explanation is valid. At any rate,

reluctantly, we leave the examination of the unexplained discount of private debt open for

further research.

7. Additional Analyses and Robustness Tests

7.1. “Transitioning” Firms

34

In addition to our sample of public and private firms (described in section 5.1

below), we also have sample firms that during the sample period have changed their

ownership from being public to being private or vice versa. While the type of ownership

of these firms in any given year is unambiguous, their characteristics as privately or

publicly owned are likely to have gradually changed in the years surrounding the

ownership transition. To strengthen the power of our tests, we exclude these “hybrid”

firms from our analyses. We take advantage of this sample of transitioning firms by

further testing the effect of ownership on the cost of debt using a difference-in-

differences design whereby we conduct “before and after” tests on the sample of the

transitioning firms. The results are reported in Table 10. These results which show that

the cost of debt of private companies, measured by either yield spreads or ratings, is

higher for private firms, after controlling for other determinants of that cost, supplement

and reinforce the main results that are based on cross-sectional tests.

7.2. Robustness tests

Non-linear numerical scale for rating: The variable RATING is based on a

conversion of S&P ratings (with AAA being the highest rating) to a numerical scale

(from 1, the highest rating, to 21). While this scale has been used by past research (e.g.,

Amato and Furfine 2004, Kisgen 2006), we conducted our main tests using a number of

alternative scales that include non-linear scale (specifically natural logarithm of, and

squaring of, the numerical rating) as well as partitioning the range between 1 and 21 to

four groups, or quartiles. The results, and in particular the sign and significance of

PRIVATE, remain intact.

35

Alternative default outcome specifications: The literature on default prediction

employing hazard rate models employs a variety of control variables as well as different

definitions of these variables.25 To ensure that our results are robust to inclusion and

definition of control variables in the hazard rate model, we conduct our hazard rate

default outcome tests using alternative subsets of variables. We also employ total assets,

instead of lagged assets, as deflators, and use LEV (total liabilities divided by total assets)

instead of LEV_ADJ. Our results are qualitatively the same under all of alternative

specifications we examined.

8. Summary and Concluding Remarks

We provide evidence in this paper that, after controlling for factors identified by past

research as affecting the cost of debt (including firm fundamentals, bond characteristics,

the firm’s information environment, as well as the endogenous nature of the choice of

being public), the cost of debt is higher for private equity firms. Specifically, after these

controls, the yield spread is higher on average by more than 1% (with the exact

percentage estimate depends on the exact test specification).

Within the privately-owned firms, we find that ownership by a PE firm results in a

higher cost of debt and higher ratings. This result is attributed by us to the greater

separation of ownership and control in PE-backed private companies, which likely leads

to more risk taking and thus a higher cost of debt. This cost-increasing effect is more

pronounced for private companies in which the PE firm holds a majority stake. However,

we also find that the positive reputational effect of PE firms is more pronounced for large

25 Shumway 2001, Campbell et al. 2008, and Bharath and Shumway 2008, among others, make comparison among different hazard rate model specifications employed in default prediction.

36

PE firms, in fact more than oversetting the unfavorable effect of ownership-control

separation.

Our main explanation of the results of a higher cost of debt issued by privately-owned

companies is their more limited access to capital markets. This explanation is supported

by two other findings of the paper. One finding is that, given their fundamentals, the

frequency of distress events is higher for private firms, a possible reflection of these

companies’ difficulty in accessing the capital markets in case of distress. The other

finding consistent with the ‘accessibility’ explanation is that the excess cost of debt to

these firms is greater at times of recession when financial distress is more common and

accessibility to capital markets becomes more critical.

The accessibility explanation cannot fully explain the excess of cost of debt for

private firms because rating agencies should properly incorporate this factor in their rate

determination. Yet, we find that the excess yield spread, albeit smaller, still exists (the

variable PRIVATE is positive and significant) even when we add to the spread regression

the bond rating as another explanatory variable. This is a somewhat surprising result that

we are unable to satisfactorily explain. It can be interpreted by either insufficient controls

for risk in our (fairly comprehensive) model or by irrational ‘bashing” of private debt by

investors. We reluctantly leave the resolution of this puzzle to future research.

The finding that the net favorable effect of being public on the cost of equity capital

extends to the cost of debt contributes to our understanding of the consequences of the

ownership choice on the firm’s cost of capital. The analyses further provide additional

insights into the role of “soft information” in the determination of credit ratings.

37

References

Acharya, V., S.T. Bharath, and A. Srinivsan. 2007. Does Industry-Wide Distress Affect Default Firms? Evidence from Creditor Recoveries. Journal of Financial Economics 85: 787-821.

Altman, E.I. 1968. Financial Ratios, Discriminant Analysis, and the Prediction of Corporate Bankruptcy. Journal of Finance 23: 589–609.

Amato, J.D. and C. H. Furfine. 2004. Are Credit Ratings Procyclical? Journal of Banking and Finance 28: 2641-2677.

Ashbaugh-Skaife, H., D. Collins, and R. LaFond. 2006. The Effects of Corporate Governance on Firms’ Credit Ratings. Journal of Accounting and Economics 42: 203-243.

Aslan, H., and P. Kumar. 2012. Strategic Ownership Structure and the Cost of Debt. Review of Financial Studies, 25(7): 2257-2299.

Badertscher, B.A., B. Jorgensen, S.P. Katz and W. Kinney. 2014. Public Equity and Audit Pricing in the U.S. Journal of Accounting Research 52 (2-3): 303-339.

Badertscher, B.A., S.P. Katz and S.O Rego. 2013. The Separation of Ownership and Control and Corporate Tax Avoidance. Journal of Accounting and Economics 56 (2-3): 228-250.

Ball, R.and L. Shivakumar. 2005. Earnings Quality in U.K. Private Firms: Comparative Loss Recognition Timeliness. Journal of Accounting and Economics 39(1): 83-128.

Bharath, S., and T. Shumway. 2008. Forecasting Default with the Merton Distance to Default Model. Review of Financial Studies 21(3): 1339-1369.

Bruner, R.F., K.M. Eades, R.S. Harris, and R.C. Higgins. 1998. Best Practices in Estimating the Cost of Capital: Survey and Synthesis. Financial Practice and Education, Spring-Summer: 13-28.

Burgstahler D.C., H.Luzi, and C. Leuz. 2006. The Importance of Reporting Incentives: Earnings Management in European Private and Public Firms. The Accounting Review 81 (5): 983-1016.

Campbell, J.Y., J. Hilscher, and J. Szilagyi. 2008. In Search of Distress Risk. Journal of Finance 63 (6): 2899-2939.

Campello, M., J. R. Graham, and C. R. Harvey. 2010. The Real Effects of Financial Constraints: Evidence from a Financial Crisis. Journal of Financial Economics 97: 470-487.

38

Claessens S., S. Djankov, J.P.H. Fan, and L.H.P. Lang. 2002. Disentangling the Incentive and Entrenchment Effects of Large Shareholdings. The Journal of Finance 57 (6): 2741-2771.

Comment, R. 2012. Revisiting the Illiquidity Discount for Private Companies: A New (and “Skeptical”) Restricted-Stock Study. Journal of Applied Corporate Finance 24 (1): 80-91.

Cooper, R. A., and J.M. Shulman. 1994. The Year-End Effect in Junk Bond Prices. Financial Analysts Journal 50 (5): 61-65.

Cox, D.R. 1972. Regression Models and Life-Tables. Journal of the Royal Statistical Society. Series B (Methodology) 34 (2): 187-220.

Das, S. R., M. Jagannathan and A. Sarin. 2002 The Private Equity Discount: An Empirical Examination of the Exit of Venture Backed Companies. Working paper. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=298083

De Fontenay, E. 2013. Private Equity Firms as Gatekeepers. Working paper. Harvard Law School. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2245156

Demiroglu, C., and C. James. 2010. The Information Content of Bank Loan Covenants. Review of Financial Studies 23 (10): 3700-3737. Donovan, J., R.M. Frankel, and X. Martin. 2013. Accounting Conservatism and Creditor Recovery Rate. Working paper. Available at SSRN: http://ssrn.com/abstract=2205968 orhttp://dx.doi.org/10.2139/ssrn.2205968

Dudney, D., and J. Geppert. 2008. Do Tax-Exempt Yields Adjust Slowly to Substantial Changes in Taxable Yields? Journal of Futures Markets 28(8): 763-789.

Erel, I., B. Julio, W. Kim, and M.S. Weisbach. 2012. Macroeconomic Conditions and Capital Raising. Review of Financial Studies 25 (2): 341-376.

Fama, E. F. 1980, Agency Problems and the Theory of the Firm, Journal of Political Economy 88: 288-307.

Fama, E., and M. Jensen. 1983. Separation of Ownership and Control. Journal of Law and Economics 26 (2): 301-325. (Corporations and Private Property: A Conference Sponsored by the Hoover Institution).

Franks, J.R., and W.N.Torous. 1994. An Empirical Investigation of U.S. Firms in Reorganization. Journal of Finance 44 (3): 747–69.

Gay, G. D., and S. Manaster. 1991. Equilibrium Treasury Bond Futures Pricing in the Presence of Implicit Delivery Options. Journal of Futures Markets 11 (5): 623-645.

39

Gilson, S., 2006. Controlling Shareholders and Corporate Governance: Complicating the Comparative Taxonomy. Harvard Law Review 119:1641–83.

Gilson, S., K. John, and L. Lang. 1990. Troubled Debt Restructurings. Journal of Financial Economics 27: 315-353.

Givoly, D. C.K. Hayn, and S.P. Katz, 2010. Does Public Ownership of Equity Improve Earnings Quality? The Accounting Review (1): 195-225.

Givoly, D. C.K. Hayn, and S.P. Katz, 2013. The Changing Relevance of Accounting Numbers to Debt Holdrs over Time. Working Paper

Grossman, S., and O. Hart. 1988. One-Share, One-Vote, and the Market for Corporate Control. Journal of Financial Economics 20: 175-202.

Hancock, D., and M.L. Kwast. 2001. Using Subordinated Debt to Monitor Bank Holding Companies: Is it Feasible? Journal of Financial Services Research 20 (2/3): 147-187.

Hand, J. R.M., R.W. Holthausen, and R.W. Leftwich. 1992. The Effect of Bond Rating Agency Announcements on Bond and Stock Prices. The Journal of Finance 47 (2): 733-752.

Harris, M., and A. Raviv. 1988. Corporate Governance: Voting Rights and Majority Rules. Journal of Financial Economics 20: 203-235.

Hemler, M. L. 1990. The Quality Delivery Option in Treasury Bond Futures Contracts. Journal of Finance 45 (5): 1565-1586.

Hertzel, M. and R. L. Smith. 1993. Market Discounts and Shareholder Gains for Placing Equity Privately. Journal of Finance 48 (2): 459-485. Ivashina, V., and A. Kovner. 2011. The Private Equity Advantage: Leveraged Buyout Firms and Relationship Banking. Review of Financial Studies 24 (7): 2462-2498. Ivashina, V., and D. Scharfstein. 2010. Bank Lending During the Financial Crisis of 2008. Journal of Financial Economics 97 (3): 319–338. Jensen, M. C. and W. H. Meckling, 1976. Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure. Journal of Financial Economics 3: 305-360. Jiang, G., C. Lee, and H. Yue, 2010. Tunneling Through Intercorporate Loans: The China Experience. Journal of Financial Economics 98:1–20.

Johnson, S., R. La Porta, F. Lopez-de-Silanes, and A. Shleifer. 2000. Tunneling. American Economic Review (Papers and Proceedings) 90: 22–27.

Kaplan, S. N. and P. Stromberg. 2009. Leveraged Buyouts and Private Equity. Journal of Economic Perspectives 23 (2): 121-146.

40

Katz, S. 2009. Earnings Quality and Ownership Structure: The Role of Private Equity Sponsors. The Accounting Review 84 (3): 623-658. Kisgen, D.J. 2006. Credit Rating and Capital Structure. Journal of Finance 61 (3): 1035- 1072.

Koeplin, J., A. Sarin, and A. Shapiro. 2000. The Private Company Discount. Journal of Applied Corporate Finance 12 (4): 94-101. Kooli, Maher, M. Kortas, and J.F. L’Her. 2003. A New Examination of the Private Company Discount: The Acquisition Approach. The Journal of Private Equity, Summer: 48-55. Kovner, A. and C. Wei. 2012. The Private Premium in Public Bonds. Working Paper, Federal Reserve Bank of New York. Lee, H. 2014. Creditor Coordination Effects and Distress Prediction. Working Paper, University of Maryland. Lin, C., Y. Ma, and Y. Xuan. 2011. Ownership Structure and Financial Constraints: Evidence from a Structural Estimation. Journal of Financial Economics 102: 416-431. Manne, H. G. 1965. Mergers and the Market for Corporate Control. Journal of Political Economy 73: 110-120. Mansi, A.S., W.F. Maxwell, and D. Miller. 2004. Does Auditor Quality and Tenure Matter to Investors? Evidence from the Bond Market. Journal of Accounting Research 42 (4): 755-793.

Masulis, R., and R. Thomas. 2009. Does Private Equity Create Wealth? The effects of Private Equity and Derivatives on Corporate Governance. University of Chicago Law Review 76 (1): 219-259.

Martin. K. and J. J. McConnell. 1991. Corporate Performance, Corporate Takeovers and Management Turnover. Journal of Finance 46 (2): 671-687.

McConnell J. J, and H. Servaes. 1990. Additional Evidence on Equity Ownership and Corporate Value. Journal of Financial Economics 27: 595-612.

Pagano, M., F. Panetta, and L. Zingales. 1998. Why Do Companies Go Public? An Empirical Analysis. The Journal of Finance 53 (1): 27-64. Shapley, S.L. 1953, A Value for n-person Games” In Contributions to the Theory of Games, volume II, by H.W. Kuhn and A.W. Tucker, editors. Annals of Mathematical Studies v. 28: pp. 307–317. Princeton University Press.

41

Shumway, T. 2001. Forecasting Bankruptcy More Accurately: A Simple Hazard Model. The Journal of Business 74 (1): 101-124. Shulman, J., and M. Bayless. 1993. Marketability and Default Influences on the Yield Premia of Speculative-Grade Debt. The Journal of the Financial Management Association 22 (3): 132-141.

Saunders, S. and S. Steffen. 2011. The Costs of Being Private: Evidence from the Loan Market. The Review of Financial Studies 24 (12): 4091- 4122.

Stein, J. C., 2003. Agency, Information and Corporate Investment, in: G.M. Constantinides, M. Harris and R. Stulz, eds.: Handbook of the Economics of Finance (Elsevier, Amsterdam).

Zmijewski, E,M. 1984. Methodological Issues Related to the Estimation of Financial Distress Prediction Models. Journal of Accounting Research 22: 59–82.

42

Appendix A Definitions of Variables (in Alphabetical Order)

Variable Definition Sources BIGAUD An indicator variable that is equal to 1 if the firm’s auditor belongs to one of the Big-

5 auditors (Big-4 after the demise of Arthur Anderson in 2002) and 0 otherwise..

COMPUSTAT

EBITDA Earnings Before Interest, Taxes , Depreciation and Amortization divided by lagged Total Assets

COMPUSTAT

DURATION Average duration of the firm’s outstanding bonds expressed as a number of years Interactive Data, TRACE

FCF Free cash flow ((Cash Flow from Operations - Average CAPEX for years t, t-1, and t-2) / Lagged Total Assets)

COMPUSTAT

FSIZE Size of the firm proxied by the natural log of the sum of the market value of equity (PRCC*SHO) and book value of debt (DT). For private companies we estimate the hypothetical market value of by multiplying the firm’s sale, and separately the firm’s assets, by the median of, respectively, the median sales multiplier and the median assets multiplier in the firm’s 3-digit industry. We then average these two market value estimates.

COMPUSTAT

GROWTH Sales Growth ((Total sales - Lagged total sales) / Lagged total sales) COMPUSTAT

INT_COV Interest coverage ratio (Income before Tax and before Interest Expense / Interest Expense, The value is set to zero when the numerator is negative

COMPUSTAT

INTANG Intangible assets (Intangible Assets / Total Assets) COMPUSTAT

LEV Leverage, computed as the ratio (Total :Liabilities / Total Assets)

COMPUSTAT

LEV_ADJ Leveraged adjusted by off-Balance Sheet adjustments , computed as [Total Liabilities + Underfunded Pension Obligations + (Rent Expense *6*.65)]/[Total Assets + .35*Underfunded Pension + LIFO Reserve *.65 + Interest Capitalized * .65 + (Rent Expense *6*.65)]

Moody (2007)

43

Variable Definition Sources LOSS An indicator variable capturing whether the firm is a loss firm (1) or a not (0) COMPUSTAT

OFFER Amount (in millions) for all issued bonds for that fiscal year. If a firm issued $100 in 2001 and $150 in 2002, OFFER would be $100 in 2001 and $250 in 2002.

FISD; SDC

PE An indicator variable representing whether the firm is owned by a private equity (1) or not (0)

Thomson Financial VentureXpert

PE_RANK An indicator variable representing whether the PE firm (PE=1) is a large PE firm (1) or not (0); Top 15 PE funds by equity invested: Carlyle Group, Blackstone, Warburg Pincus, KKR, Goldman Sachs, Cerberus Capital, Fortress, Apollo, Bain, TPG, 3i, Apax, Thomas and Lee, Morgan Stanley, Welsh, Carson, Anderson & Stowe.

Thomson Financial VentureXpert

PE_MAJ An indicator variable representing whether the PE firm in the PE-owned private firm has a majority interest (≥ 50%) in the firm (1) or not (0).

Hand-collected

PPE Gross property, plant, and equipment (Total gross property plant and equipment / Lagged total assets)

COMPUSTAT

PRIVATE An indicator variable representing private (1) vs. public (0) status of the firm Hand-collected

PUBLIC_NEW An indicator variable equal to 1 if the firm-year is less than 5 years from the year the firm became public (by issuing to the public either debt or equity)

COMPUSTAT

RATING S&P credit rating converted into numeric ratings with the highest rating (AAA) is assigned the value of 1 and the lowest rating is assigned the value 21. Measured at the end of the third month of the fiscal year.

COMPUSTAT

RE Retained Earnings (RE) divided by total assets COMPUSTAT

REC An indicator variable equal to 1 if the fiscal year is a contraction (i.e., recession) year according to the NBER website http://www.nber.org/cycles/cyclesmain.html. The fiscal years included are 1990, 1991, 2001, 2008, and 2009.

NBER website

44

Variable Definition Sources ROA Return on assets, computed as the ratio of (Income from Continuing Operations

before Extraordinary Income / Lagged Total Assets)

COMPUSTAT

SALES_TURN Sales divided by Total Assets COMPUSTAT

SENIORITY An indicator variable that is set equal to one if the firm has a senior bond, and zero otherwise.

FISD; SDC

SPREAD The excess of yield-to-maturity of the bond on a matched Treasury bond that has at the end of the third month of the fiscal year (1) a remaining time to maturity within six months from the remaining time to maturity of the sample bond, and (2) a coupon rate closest to that of the sample bonds but one that does not deviates from the coupon rate of the sample bond by more than 45%.

Interactive Data, TRACE, CRSP

WC Working Capital (Current Assets – Current Liabilities) divided by lagged Total Assets.

COMPUSTAT

YRS_TO_MATURITY

Remaining number of years to maturity.

Interactive Data, TRACE

YTM Yield to Maturity, measured at the end of the third month of the fiscal year.

Interactive Data, TRACE

Z-SCORE An updated Altman's Z from Shumway (2001) where the weights on the coefficients change: 1.2*X1+0.6*X2+10.1*X3-0.47*X5 [data 1962-1992] [see Table 2, page 117 of Shumway 2001].

COMPUSTAT

45

Table 1 Descriptive Statistics on the Firm Sample

Panel A: Distribution of firm-year observations by year

YEAR Number of

Private Firms Number of

Public Firms

1987 11 976 1988 20 914 1989 24 841 1990 23 789 1991 20 813 1992 22 876 1993 19 952 1994 25 997 1995 25 1,108 1996 23 1,248 1997 38 1,386 1998 54 1,492 1999 84 1,510 2000 103 1,526 2001 127 1,489 2002 114 1,440 2003 100 1,465 2004 81 1,457 2005 63 1,419 2006 49 1,395 2007 40 1,324 2008 33 1,264 2009 29 1,248 2010 23 1,264

1,150 29,193

46

Table 1 Panel B: Distribution of firm observations by Industry*

Industry (2-digit

SIC code)

Number of

Private Firms

% of Private Firms

Number of Public

Firms

% of Public Firms

% of All Sample Firms

% of all COMPUSTAT

Firms

(1) (2) (3) (4) (5) (6)

13 2 0.8% 210 6.1% 5.8% 7.4% 20 12 4.7% 101 3.0% 3.1% 2.4% 26 9 3.5% 75 2.2% 2.3% 0.9% 27 14 5.5% 57 1.7% 1.9% 0.8% 28 18 7.0% 209 6.1% 6.2% 10.7% 29 4 1.6% 54 1.6% 1.6% 0.8% 33 7 2.7% 89 2.6% 2.6% 1.4% 34 10 3.9% 50 1.5% 1.6% 1.1% 35 16 6.3% 154 4.5% 4.6% 4.6% 36 9 3.5% 183 5.4% 5.2% 8.2% 37 12 4.7% 103 3.0% 3.1% 2.2% 38 7 2.7% 106 3.1% 3.1% 5.5% 48 10 3.9% 346 10.1% 9.7% 3.7% 49 11 4.3% 290 8.5% 8.2% 3.8% 50 10 3.9% 65 1.9% 2.0% 1.9% 51 5 2.0% 53 1.6% 1.6% 1.3% 54 9 3.5% 51 1.5% 1.6% 0.5% 58 7 2.7% 50 1.5% 1.6% 1.1% 59 5 2.0% 67 2.0% 2.0% 1.5% 73 14 5.5% 222 6.5% 6.4% 11.2% 79 11 4.3% 60 1.8% 1.9% 0.9% 80 6 2.3% 87 2.5% 2.5% 1.4%

All other industries

48 18.8% 733 21.5% 21.3% 26.7%

Full sample

256 100.0% 3,415 100.0% 100% 100.0%

*Listed are industries with at least 50 public firms in our sample.

47

Table 1 Panel C: Descriptive Statistics on Firm and Bond Characteristics

Firm Characteristics

Number of firm-year with available

SPREAD or RATING Mean Median std.dev.

Number of firm-year with available

SPREAD or RATING Mean Median std.dev. Mean MedianLEV_ADJ 1,150 1.053 0.988 0.293 29,193 0.728 0.689 0.259 0.326 *** 0.299 ***INT_COV 1,150 1.328 1.138 1.073 29,193 4.147 3.119 3.160 -2.819 *** -1.982 ***Z_SCORE 1,150 0.265 0.327 0.779 29,193 0.592 0.612 0.958 -0.326 *** -0.285 ***ROA 1,150 -0.014 -0.003 0.083 29,193 0.027 0.038 0.093 -0.042 *** -0.041 ***LOSS 1,150 0.519 1.000 0.500 29,193 0.234 0.000 0.423 0.285 *** 1.000 ***FCF 1,150 0.017 0.018 0.087 29,193 0.036 0.028 0.197 -0.019 *** -0.010 ***INTANG 1,150 0.250 0.173 0.255 29,193 0.152 0.062 0.207 0.098 *** 0.111 ***PPE 1,150 0.323 0.265 0.231 29,193 0.433 0.387 0.281 -0.110 *** -0.122 ***GROWTH 1,150 -0.090 0.034 0.691 29,193 0.025 0.044 0.448 -0.115 *** -0.010 FSIZE 1,150 7.085 6.939 1.001 29,193 7.898 7.835 1.624 -0.813 *** -0.896 ***BIGAUDI 1,150 0.936 1.000 0.245 29,193 0.957 1.000 0.202 -0.022 *** 0.000 PUBLIC_NEW 1,150 0.530 1.000 0.499 29,193 0.160 0.000 0.367 0.370 *** 1.000 ***

Bond Characteristics Number of firm-years Mean Median std.dev. Number of firm-years Mean Median std.dev. Mean Median BONDS 634 1.240 1.000 0.293 11,389 3.160 2.000 4.458 -1.920 *** -1.000 ***SPREAD (%) 634 9.579 7.937 1.073 11,389 3.333 1.935 3.860 6.246 *** 6.002 ***MATURITY 634 9.361 10.000 0.779 11,389 13.478 10.000 8.026 -4.116 *** 0.000 YRS_TO_MATURITY 634 5.518 5.625 0.083 11,389 9.187 7.000 7.365 -3.669 *** -1.375 **OFFER 634 5.025 4.832 0.500 11,389 5.913 5.720 1.260 -0.888 *** -0.889 ***DURATION 634 3.764 3.760 0.087 11,389 5.263 5.244 2.231 -1.499 *** -1.484 ***SENIORITY 634 0.303 0.000 0.255 11,389 0.731 1.000 0.407 -0.429 *** -1.000 ***RATING (from 1 to 21) 1,015 14.515 14.000 0.231 28,433 10.191 10.000 3.851 4.324 *** 4.000 ***

DifferencePrivate Firm-Years Public Firm-Years

*,**,*** indicate significance at the 10%, 5%, and 1% level using a two-tailed t-test, respectively. Differences between means are tested for significance using a two-tailed t-test; differences in medians are tested for significance using a two-tailed Wilcoxon signed rank test. All variables are as defined in the Appendix.

48

Table 2: Determinants of Cost of Debt: Results from Regression (1) – Full Sample

SPREAD RATING

Coeff t-stat Coeff t-stat

Intercept 10.959 24.50 19.013 80.84

PRIVATE 2.777 10.73 0.851 8.99

LEV_ADJ 1.277 5.71 2.121 20.62

INT_COV -0.067 -2.81 -0.222 -19.56

Z_SCORE -0.190 -1.89 -0.531 -11.58

ROA -3.776 -3.55 -1.015 -1.71

LOSS 1.680 12.51 1.096 13.87

FCF -1.077 -4.73 0.037 0.19

INTANG -0.865 -4.09 0.602 5.14

PPE -0.763 -4.82 -0.661 -6.54

GROWTH -0.248 -2.00 0.313 3.46

DURATION -0.201 -10.93 -0.209 -18.36

SENIORITY -0.917 -7.38 -1.114 -17.55

FSIZE -0.703 -18.08 -0.922 -46.60

BIGAUDI -0.088 -0.32 -0.063 -0.38

PUBLIC_NEW 0.311 2.16 0.626 7.75

Industry FE Yes Yes

Year FE Yes Yes

Adjusted R-square 54.23% 62.86%

N 12,023 11,128

*,**,*** indicates significance at the 10%, 5%, and 1% level, respectively. All variables are as defined in Appendix A. Regressions include industry and year indicator variables, which have not been tabulated. The t-stats have been adjusted to control for the clustering

by year and multiple firm observations.

49

Table 3: Results from Estimating the Propensity Score Regression

Panel A: Regression Estimates

Intercept LOSS PPE INTANG GROWTH INV_COV MacKelvey Pseudo

R2 Coefficient -1.362 -0.074 -0.801 1.312 -0.189 -0.711 0.418 Odds Ratio na 0.928 0.449 3.712 0.828 0.491 z-statistic -13.79 -1.03 -9.39 -10.57 -3.25 -22.36

50

Table 3: Results from Estimating the Propensity Score Regression

Panel B: Main Characteristics of the Matched Observations

Firm Characteristics

Number of firm-year with available SPREAD

or RATING Mean Median std.dev.

Number of firm-year with available SPREAD

or RATING Mean Median std.dev. Mean Median LEV_ADJ 343 1.090 1.018 0.306 343 1.056 0.977 0.283 0.034 0.041 INT_COV 343 1.359 1.178 0.838 343 1.977 1.436 1.940 -0.618 *** -0.259 ***Z_SCORE 343 0.250 0.257 0.715 343 0.175 0.211 1.147 0.074 0.046 ROA 343 -0.009 0.002 0.076 343 -0.015 0.003 0.110 0.006 -0.001 LOSS 343 0.481 0.000 0.500 343 0.472 0.000 0.500 0.009 0.000 FCF 343 0.018 0.020 0.076 343 0.001 0.013 0.102 0.017 ** 0.007 *INTANG 343 0.217 0.155 0.229 343 0.209 0.116 0.255 0.008 0.039 PPE 343 0.371 0.303 0.248 343 0.409 0.348 0.277 -0.038 *** -0.045 ***GROWTH 343 0.012 0.048 0.595 343 0.066 0.048 0.382 -0.054 0.000 FSIZE 343 6.930 6.859 0.927 343 6.815 6.835 1.086 0.114 0.024 BIGAUDI 343 0.948 1.000 0.223 343 0.948 1.000 0.223 0.000 0.000PUBLIC_NEW 343 0.554 1.000 0.498 343 0.143 0.000 0.350 0.411 *** 1.000 ***

Bond Characteristics Number of firm-years Mean Median std.dev. Number of firm-years Mean Median std.dev. Mean Median BONDS 167 1.281 1.000 0.828 167 1.830 1.000 2.086 -0.549 *** 0.000 SPREAD (%) 167 9.062 7.641 5.767 167 6.448 5.139 4.893 2.614 *** 2.502 ***MATURITY 167 9.460 10.000 2.152 167 10.504 10.000 4.678 -1.044 ** 0.000 YRS_TO_MATURITY 167 5.735 5.667 2.267 167 6.544 5.583 4.386 -0.809 ** 0.083 OFFER 167 1.608 1.590 0.112 167 1.692 1.667 0.146 -0.084 *** -0.077 ***DURATION 167 3.873 3.839 1.095 167 4.606 5.094 2.072 -0.733 *** -1.255 ***SENIORITY 167 0.266 0.000 0.417 167 0.443 0.333 0.467 -0.177 *** -0.333 ***RATING (from 1 to 21) 343 14.504 14.000 1.837 343 13.770 14.000 2.943 0.735 *** 0.000

DifferencePrivate Firm-Years Public Firm-Years

*,**,*** indicate significance at the 10%, 5%, and 1% level using a two-tailed t-test, respectively. Differences between means are tested for significance using a two-tailed t-test; differences in medians are tested for significance using a two-tailed Wilcoxon signed rank test. All variables are as defined in the Appendix.

51

Table 4: Determinants of Cost of Debt: Results from Regression (1) – Propensity Matched Sample

SPREAD RATING

Coeff t-stat Coeff t-stat

Intercept 24.014 5.453 24.355 14.47 PRIVATE 2.481 4.336 0.444 1.75 LEV_ADJ 2.391 3.062 1.567 4.61 INT_COV 0.002 0.008 -0.355 -3.07 Z_SCORE -0.837 -2.009 -0.905 -4.12 ROA -4.586 -1.052 -0.325 -0.16 LOSS 1.394 1.906 -0.394 -1.30 FCF -6.094 -2.100 0.203 0.14 INTANG -1.154 -1.057 -0.021 -0.05 PPE -2.068 -2.081 -1.034 -2.19 GROWTH -0.367 -0.446 1.087 3.03 DURATION -0.392 -2.635 -1.087 -6.28 SENIORITY -0.692 -1.340 -0.282 -0.83 FSIZE -1.003 -3.547 0.261 1.10 BIGAUDI -0.831 -0.899 -0.150 -2.22 PUBLIC_NEW -0.274 -0.398 -0.897 -3.49

Industry FE Yes Yes Year FE Yes Yes Adjusted R-square 60.39% 64.79% N 334 334

*,**,*** indicates significance at the 10%, 5%, and 1% level, respectively. All variables are as defined in Appendix A. Regressions include industry and year indicator variables, which have not been tabulated. The t-stats have been adjusted to control for the clustering by year and multiple firm observations.

52

Table 5 Effect of Recessionary Years on the Cost of Debt

SPREAD RATING

Coeff t-stat Coeff t-stat

Intercept 10.906 24.12 19.069 81.16 PRIVATE 2.239 9.32 0.753 7.26 REC 0.120 0.94 0.445 5.90 PRIVATE*REC 2.688 5.85 0.627 2.96 LEV_ADJ 1.276 5.71 2.096 20.38 INT_COV -0.068 -2.86 -0.225 -19.81 Z_SCORE -0.188 -1.87 -0.525 -11.45 ROA -3.831 -3.61 -1.095 -1.83 LOSS 1.687 12.60 1.107 13.99 FCF -1.082 -4.75 0.074 0.39 INTANG -0.870 -4.13 0.584 5.00 PPE -0.769 -4.86 -0.667 -6.60 GROWTH -0.208 -1.67 0.286 3.16 DURATION -0.201 -10.98 -0.209 -18.41 SENIORITY -0.911 -7.35 -1.130 -17.81 FSIZE -0.702 -18.07 -0.917 -46.33 BIGAUDI -0.107 -0.40 -0.045 -0.27 PUBLIC_NEW 0.304 2.11 0.604 7.48

Industry FE Yes Yes Year FE Yes Yes Adjusted R-square 54.55% 62.99% N 12,023 11,128

*,**,*** indicates significance at the 10%, 5%, and 1% level, respectively. All variables are as defined in Appendix A. Regressions include industry and year indicator variables, which have not been tabulated. The t-stats have been adjusted to control for the clustering by year and multiple firm observations.

REC=1 if year is equal to a recessionary year defined by NBER (1990, 1991, 2001, 2008, 2009).

53

Table 6

The Effect of PE Ownership and PE Characteristics on the Cost of Debt

All Private Companies Private Companies with Private Equity Ownership

DUMMY = PE DUMMY = PE_MAJ DUMMY = PE_RANK

SPREAD RATING SPREAD RATING SPREAD RATING

Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat

Intercept 27.787 11.403 20.686 41.148 27.815 10.285 19.267 37.942 29.271 11.155 19.192 40.547

DUMMY 0.312 1.875 0.336 2.800 0.873 1.803 0.070 0.465 -1.331 -2.976 -0.404 -3.103 LEV_ADJ 0.722 1.140 1.041 6.184 0.919 1.277 0.923 4.960 0.824 1.172 0.904 4.822 INT_COV -1.176 -4.471 -0.568 -9.262 -2.214 -3.475 -0.309 -3.372 -2.089 -3.346 -0.311 -3.385 Z_SCORE -0.239 -0.510 -0.483 -5.746 0.279 0.450 -0.460 -4.499 0.269 0.433 -0.475 -4.594 ROA -4.329 -1.294 -1.281 -1.369 -8.708 -2.518 -1.349 -1.411 -9.483 -2.798 -1.268 -1.322 LOSS 1.523 3.217 -0.035 -0.307 -0.040 -0.069 -0.012 -0.098 0.048 0.083 -0.013 -0.110 FCF -3.430 -1.157 0.105 0.238 -4.568 -1.093 -0.717 -1.716 -3.799 -0.914 -0.726 -1.751 INTANG -3.720 -5.059 -0.805 -4.505 -2.882 -3.184 -0.703 -3.888 -2.504 -2.827 -0.718 -4.105 PPE -3.024 -3.964 -1.502 -5.098 -0.927 -0.753 -1.776 -4.051 -1.335 -1.099 -1.752 -4.196 GROWTH -0.446 -1.322 0.107 1.341 -0.031 -0.068 -0.035 -0.481 -0.012 -0.028 -0.034 -0.468 FSIZE -0.631 -3.151 -0.576 -9.314 -0.523 -1.982 -0.386 -6.617 -0.543 -2.092 -0.394 -6.797 BIGAUDI -0.252 -0.398 -0.570 -3.148 -0.450 -0.498 0.101 0.680 -0.824 -0.827 0.131 0.870 PUBLIC_NEW -0.498 -1.331 -0.052 -0.573 -0.250 -0.522 -0.223 -2.068 -0.117 -0.244 -0.237 -2.301

Industry FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes

Adjusted R2 59.90% 51.75% 59.73% 46.28% 60.18% 46.96%

N 633 1,014 415 703 415 703 *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively. All variables are as defined in Appendix A. Regressions include industry and year

indicator variables, which have not been tabulated. The t-stats have been adjusted to control for the clustering by year and multiple firm observations.

54

Table 7: The Effect of Private Ownership on Distress Probability

Coef z-stat P-values Intercept -2.552 -9.459 0.000 *** PRIVATE 0.326 2.544 0.011 ** LEV_ADJ 0.267 1.882 0.060 * INT_COV -0.338 -11.368 0.000 *** ZSCORE -0.499 -8.662 0.000 *** ROA -3.975 -8.415 0.000 *** LOSS 0.941 6.873 0.000 *** FCF -1.348 -3.699 0.000 *** INTANG -0.435 -2.073 0.038 ** PPE -0.617 -3.488 0.000 *** GROWTH 0.186 2.147 0.032 ** FSIZE -0.162 -5.274 0.000 ***

Pseudo R2 28.3%No of obs* 30,877

*The sample consists of 2,094 observations belonging to private companies (of which 111, or 5.3% are distress observations) and 29,142 observations belonging to public companies (of which 627, or 2.2% are distress observations).

55

Table 8: The Effect of Private Ownership on Recovery Rates of Debt

Coef P-value

Intercept 52.260 0.024 ** PRIVATE 0.958 0.821 EBITDA 32.015 0.010 *** TA 4.174 0.001 *** TAN -0.299 0.950LEV_ADJ -9.232 0.047 ** DUR -0.255 0.019 ** DE 17.101 0.000 *** SECURED 3.080 0.274

Industry Fixed Effects Yes Year Fixed Effects Yes Adjusted R-Squared 23.5%No of Obs. 333

*The sample consists of 374 of the 738 distress observations for which recovery data are available. Further data requirements for estimating regression (3) reduced this sample to 333 observations, out of which 56 are of private companies and 318 of public companies.

56

Table 9: The Effect of Private Ownership on the Yield Spread after Controlling for Rating

   Full Sample Propensity-

Matched Sample    Coeff t-stat

Intercept 2.320 4.619 6.131 0.921 PRIVATE 1.468 9.321 1.655 3.869 LEV_ADJ 0.331 1.756 1.240 1.594 INT_COV 0.057 2.971 0.262 1.298 Z_SCORE -0.082 -1.061 -0.173 -0.414 ROA -2.743 -2.587 -4.347 -1.128 LOSS 1.259 9.681 1.683 2.456 FCF -0.995 -3.434 -6.243 -2.200 INTANG -0.657 -3.392 -1.138 -1.117 PPE -0.506 -3.525 -1.309 -1.419 GROWTH -0.374 -2.808 -1.165 -1.546 DURATION -0.246 -6.671 -0.205 -0.601 SENIORITY -0.315 -1.536 -0.624 -0.696 FSIZE 0.100 0.686 -0.466 -0.716 BIGAUDI -0.136 -7.527 -0.281 -2.025 PUBLIC_NEW -0.538 -4.862 -0.033 -0.064 RATING 0.417 21.670 0.734 4.318

Industry FE Yes Yes Year FE Yes Yes Adjusted R-square 60.59% 64.84% N 11,128 334

*,**,*** indicates significance at the 10%, 5%, and 1% level, respectively. All variables are as defined in Appendix A. Regressions include industry and year indicator variables, which have not been tabulated. The t-stats have been

adjusted to control for the clustering by year and multiple firm observations.

57

Table 10 Results from the Sample of Transitioning Firms

PUBLIC to PRIVATE transition PRIVATE to PUBLIC transition

SPREAD RATING SPREAD RATING

Coeff t-stat Coeff t-stat Coeff t-stat Coeff t-stat

Intercept 10.960 5.652 20.271 30.538 16.745 5.935 24.307 23.863

PRIVATE 2.057 3.822 0.706 4.767 0.563 1.630 0.292 2.045 LEV_ADJ -0.693 -0.663 1.546 6.454 -0.150 -0.105 1.765 5.919 INT_COV -0.343 -1.945 -0.436 -9.551 0.020 0.090 -0.481 -8.634 Z_SCORE -0.935 -1.564 -0.582 -4.727 -1.552 -2.134 -0.962 -6.338 ROA -6.115 -0.778 -0.335 -0.245 -7.092 -0.790 0.973 0.580 LOSS 0.882 1.365 0.153 0.946 0.845 1.028 0.125 0.591 FCF -0.072 -0.028 1.168 2.104 -4.141 -0.968 0.559 1.112 INTANG -0.511 -0.560 -0.898 -3.136 1.086 0.486 -0.742 -2.229 PPE 0.645 0.444 -0.777 -2.163 2.158 1.510 -1.069 -2.882 GROWTH 0.388 0.739 -0.092 -0.517 -0.419 -0.502 0.138 1.026 FSIZE -0.596 -3.329 -0.713 -15.076 -0.792 -2.516 -1.124 -13.200 BIGAUDI -2.858 -3.403 -0.836 -2.204 -0.565 -0.448 0.360 0.471 PUBLIC_NEW 0.912 1.442 0.640 3.812 -1.659 -2.826 -0.064 -0.414

Industry FE Yes Yes Yes Yes Year FE Yes Yes Yes Yes

Adjusted R2 61.07% 66.40% 65.78% 72.55%

N 289 908 173 640 *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively. All variables are as defined in Appendix A. Regressions include industry and year indicator variables, which have not been tabulated. The t-stats have been adjusted to control for the clustering by year and multiple firm observations.